Vol. 4. No. 1 A-1 July 1999
Return to Table of Contents Return to Main Page

Reference in Online Discourse

Anita Pincas
Institute of Education, University of London


Effective reference to previous parts of the discourse is especially crucial in a collaborative online context, such as a distance learning course using computer-mediated communication (CMC). The paper surveys and measures the way a group of students in a recent CMC course attempted to develop referencing conventions to suit their learning purposes, relying heavily on actual quotation (citation) from previous messages. It is suggested that success in using referencing correlates with success in a CMC course and that without it CMC users find it difficult to develop an effective mental representation of the discourse.


A "hot" topic in education today is whether computer mediated communication (CMC) is as successful as face to face talk (F2F) or other forms of distance education. This paper relates the debate to one crucial aspect, namely direct or indirect quotation of something already said - here called "X-citation" (cross-citation) in the asynchronous CMC context. It deals specifically with CMC users' processes of creating written X-citation equivalents for familiar oral conventions that are central to effective interaction between people, and, it is suggested, to a workable mental representation of CMC discourse.

The data suggests that because CMC course delivery depends heavily on learner discussion, well-managed X-citation can be an index of success. The paper also considers the link between establishment of good X-citation conventions and the influence of software. The discussion is based on a course using the conferencing system, FirstClass. It is a message-based system, so that comments apply equally to much other modern conferencing software and to e-mail, unless specific features of FirstClass are mentioned.

The course was the twenty-week part-time Certificate in Online Education and Training (October 1996 to February 1997). This is a course that seeks to expose educators and trainers to the student experience of online distance study. There were 50 enrolled academics and trainers from around the world, all either native speakers or competent users of English. [-1-]

The Problem

Developing a useful mental representation of written, asynchronous CMC discourse is a new and therefore difficult process for inexperienced participants. It requires a learning effort for which they may not be prepared. Many can find it a negative experience in the early stages. During the course on which the present data is based, students openly indicated their stress:

Student A.
I am feeling very lost. Having lost my connection for some time, I now cannot find Group A anywhere. -- Is it a new conference? I dont see any icon for a conference called Group A !! Help !!

Student B.
Once I had lost it, there was no way back. I made a feeble effort to climb back onto Module 3, but ... even when I could see that the quality of input and discussion was amazing -- In the end, I would just look at the FirstClass icon and refuse to click ...

Even those who followed the course until the end wished they had had some more orientation:
Student C.
I think something that would have helped me is to have a map - a plan of what was going on and when... probably on paper, or in a place in First Class where I could always go and look to it.

Such lack of orientation obviously needs to be avoided as much as possible. This paper will show that a helpful factor is clear methods of referring back to previous items in the CMC discussion.

Comparison between familiar face-to-face, here-and-now discourse (henceforth referred to as F2F), and asynchronous CMC, indicates that the key difficulties in the latter arise from problems of unfamiliar space, not seeing the other participants, and lack of predictable timeframes.

In F2F contexts, members of a discussion can normally hold the unfolding events in their minds well enough to feel that they are keeping abreast. They are helped by being in the fixed space of a room, by having a bounded timeframe, and by actually seeing the others. They have at their disposal established modes of turn-taking and many conventionalised ways of referring back to previous elements in the discourse (Saville-Troike, 1982; Boden, 1994). [-2-]

Clear topic reference, and especially reference to what previous speakers have said, is critical among the important discourse skills that speakers need for any normal discussion (Halliday & Hassan, 1976, 1990). Without it, participants could not integrate the various messages into a coherent and cohesive whole and the interaction would be felt to be random or even meaningless. Users need to bind different speakers' contributions together in order to form a mental representation of the discussion.

Topic reference is possibly the major factor in course delivery where collaborative group learning underpins the course design (McConnell 1994). In a F2F context, because of the high immediacy and great speed, there is rarely a need for much of an original message to be repeated. The interlocutors take turns, each successive turn being linked to the previous one either by simple contiguity in an adjacency pair, sometimes with brief linking devices, such as "What's more", "No, but", "Well", or - where the adjacency has been interrupted - with brief signals like "Going back to what Jack said before...", "On the point about language that Mary made,...", etc. They can safely assume that most of the preceding discussion is fresh in everyone's mind, as long as the interaction has a manageable time span and the topic has a narrow focus, as in most educational seminars. In a learning environment such collaboration is crucial. A theme is being explored and participants deal with each other's ideas.

In asynchronous CMC, the usual default order in which messages are presented - chronological - is the analogue of a F2F discussion. Why then does it create problems?

First, because writing takes place on a computer, the discussion does not occur in any place of the expected kind. Thus, one familiar framing device, which would help to create a mental representation, is lacking. Second, the other speakers are not visible, and may not even be known to each other except by name and partial information. Again, an important element for representation of the flow of utterances is absent.

Third, the timeframe is often too vaguely perceived to help. It may be known that the discussion is bounded in days or weeks, but individual participants have no real sense of when their fellow interlocutors are taking part. An individual is aware only of the block of messages available here and now in the computer. (Headers giving the time when messages were sent are too peripheral to affect this.) [-3-]

Fourth, writing has the effect of changing users' expectations. Whereas they accept that they will forget a great deal of evanescent speech, they now expect to have all the information on record at their fingertips. Ideally, they would want to be able to "flick back" to previous messages. But this is impractical in current online systems because it is not possible to provide hyperlinks for everything an individual might want, and messages in screen windows cannot be leafed or scrolled through the way text can.

All these features of CMC taken together can have an effect like hearing a whole discussion at once - it becomes "noise" and can be unnerving. A further problem is that CMC discussions may split up into many different threads. In F2F interactions dominant speakers use loudness, eye contact and body movement to channel a diffuse discussion into focus, but CMC makes dominance hard to establish.

Even print in bold, italics, underlined, capitalised, or coloured form does not have the desired effect because readers may find varied fonts irritating, and in any case modern software allows readers to alter whatever fonts a writer has used, thus (perhaps unwittingly) destroying the effect. Dominance therefore has to be established by other means, perhaps message length, message frequency, or prior status known to the participants. But even so, CMC cannot prevent all, including the non-dominant, from contributing many messages of any length. The volume of a CMC discussion is therefore often far larger than in a F2F one, imposing a greater cognitive burden on all involved (Harasim, 1989; Robinson, 1993).

Another burden is the users' need to adjust to the demands of visual processing of information. Even though they may be experienced competent readers, their screen reading is likely to be slower than normal, especially since they now need the skills to respond to visual discourse signals which may be highly idiosyncratic for different writers, such as:

- headers
- type styles
- layout
- colour

These computer-mediated signs replace the familiar paralinguistic discourse signals of F2F interaction, and add to users' attested feelings of considerable uncertainty (Howell-Richardson, 1995, 1996; Howell-Richardson & Mellar, 1996). Within a CMC discussion of more than two people, writers do not know who, if anyone, has read or will read their messages, or anyone else's, let alone respond to them. They are therefore much less certain of meaning and illocutionary force. An indication of this is that they are driven to be far more explicit than they might be in a F2F context (Howell-Richardson & Mellar, 1996). [-4-]

What the present data shows, in addition to a reasonable use of the cohesive and referring devices of ordinary language, is a strikingly high use of actual quotations from previous messages. This is quite foreign to F2F discussion. It seems that our CMC participants felt that they could not rely on their readers' memory of past messages. They seemed to be using X-citation as a toe-hold, a partial framing device. Unfortunately, as will be shown, they experienced considerable difficulty in using X-citation efficiently.

Given all the difficulties that stand in the way of achieving a coherent picture of large numbers of messages that are accumulating without a readily perceivable structure, it seems fair to say that users would be greatly assisted by an efficient means of referring back to previously said, ie written, parts of the discussion. This should be part of our goal as educators, to help them create the kind of mental representation of a discussion that would enable them to feel as comfortable as in F2F discourse.

There have been many attempts to analyse the structures and content of CMC interaction since the late 1970s (Mason & Kaye, 1990; Kaye, 1991), Rowntree, 1995; Mason, 1993; December, 1993; Paulsen, n.d.), and a very small number of research papers on classroom discourse in higher education (Catt & Eke, 1995). However, I am not aware of any research into methods of back- and cross-reference in CMC. The data on which this paper is based evidences a struggle by users to develop such conventions of reference, as well as a link between apparent success and participation in the course. That link will be presented first, followed by a discussion of how reference conventions can be developed.

Citation Use and Participation Rates

Before evaluating the evidence that efficient referencing leads to greater participation in CMC, it is necessary to place the statistics in context, in particular to note the students' very uneven participation. In order to distinguish between reference in spoken discussion, and the analogous phenomenon in CMC, which is in fact quotations of what previous discussants have written, the term "X-citation" (cross-citation) will be used for the latter. [-5-]

There were great differences in the numbers of words and messages produced by the 50 students. Within the five identifiably different activities - four modules sequenced across the 20 weeks and one "workshop" which was a parallel activity for most of the period - 4 students never posted any messages. While only just over half, 26 students, took part in at least three of the five activities, 16 took part in four, but only 12 appeared in all five.

Nor did they send messages at even rates during the period of the course. There were bursts of activity which varied from participant to participant. However, interestingly, 71% of all messages by any one student were posted within two days of that student having posted a previous message.

These aspects of variation link to two relevant aspects of CMC course delivery. First, many inactive students dropped out, and, with hindsight, it seems predictable that if more than a week went by without a message from an inactive student, s/he was likely not to reappear. Second, the active students logged on frequently and posted several messages every time they did so, responding to various points as they worked their way through the other students' messages.

But, again, there is a great deal of variation. The volume of messages was greatest during Module 1 but the volume of words was greatest during the Module 2. Overall, it is clear that participation steeply declined from Module 3 onwards, i.e., from roughly Week 13.

The number of any student's messages did not increase linearly with the number of activities s/he took part in, but those students who took part in most activities did post more messages. This would not necessarily be so; a student could participate very heavily in a small number of activities, and thus post more messages than one who participated only slightly in all of them. On this course, however, students who took part in three of the five activities sent an average of .8 messages per week. Those who took part in four or five, sent a weekly average of 1.7 and 3 messages respectively. (The tutor's initial recommendation had been to post about 2 messages a week.)

It is the background of uneven participation which enables a comparison between the more or less active students in their use of X-citation

Figure 1 gives a breakdown of the use of X-citations - from 1 to 4 or more - according to how many of the activities students participated in. Most messages in the course contained none. But the last column shows that those who took part in all five activities used the most. The students who took part in between one and three activities used X-citations in up to 12.1% of their messages. By contrast, those who participated in four activities doubled that use to 24% of their messages, while those who took part in all five trebled it to 36%. Further, Figure 1 also shows that these students used more X-citations within a single message than the others, though using more than two per message was rare. [-6-]

These differences are great enough to link X-citation usage and increased participation. If one equates fuller participation with successful participation, one could go on to deduce that the X-citation figures are an index of it. This makes intuitive sense, because one would reasonably expect X-citation to be an indicator of interaction, which, in turn, one would normally equate with successful discussion, which, in an educational context, one could equate with collaborative learning. Research on larger numbers, together with information about subjects' views of their success, might well confirm this.

The other major statistical finding to support the hypothesis relates to the quantity and size of messages. Students who took part in either four or all five activities divide into three categories - prolific, medium, and limited, according to the total number of words they wrote, while the less active students who completed only between one and three activities wrote even less and form one group called minimal writers.

prolific writers20,000to40,000 words
medium writers 10,000to20,000 words
limited writers1000to10,000 words
minimal writers <1000 words

When the weekly average number of X-citations in messages from these groups was calculated, it emerged that during the first eight weeks the use of X-citations took off at a different point for each group, as follows:

prolific writersstarted using X-citations in the 3rd week.
medium writers started using X-citations in the 5th week.
limited writersstarted using X-citations in the 7th week.
minimal writersstarted using X-citations in the 8th week.

Thus, the students who took part in most of the course, not only wrote more words and used more citations than the others (as shown in Figure 1), but started using X-citations significantly sooner. [-7-]

The least active people, the minimal writers, did make an effort to learn interactive techniques, as shown by their eventual use of X-citations in week 8, but they nevertheless failed to become active students. This is not because they were late (in fact they were already working by the second week), or because they did not write messages (they each contributed an average of 1 to 2 messages a week at the beginning). The only reliable factor in the present data that distinguishes them from the other categories is their late use of X-citations.

Taken together, the figures seem to justify the view that X-citation is an indicator of participation success. The most stable group were the limited writers who wrote an average of 1.43 message of about 150 words per week. This would seem to be sufficient to keep abreast of a 20-week CMC course, and may serve as a guide for students who worry about how much they should write and how often. Indeed, 50 students contributing at this rate would generate a very manageable amount of weekly material to read (7500 words).

Nevertheless, I recognise there are other possible explanations of the above figures. It could be that students who were motivated to play a fuller part in the course were thereby motivated to pay more attention to the messages of their peers and to respond to them, and therefore to cite them. Or, it could be the other way round: students who were more responsive to the contributions of their peers were thereby motivated to play a bigger part in the course. Or again, the two features, high use of X-citation and high participation, may simply indicate technical skill.

There is, moreoever, an interesting puzzle presented by Figure 1. If one leaves aside the active participants and considers just the cluster who took part in one to three activities, those who took part in only one or two used somewhat more X-citations than students who took part in three. Does this falsify the hypothesis correlating X-citation with greater participation? One reply might be that the "3 activities" people replaced X-citations by their own paraphrases, a feature it was not possible to include in this study, but which clearly merits further research.

But there is another factor that seems relevant to the puzzle, namely that Module 3 has

a. the greatest number of messages per thread

b. the greatest average length of threads (i.e., for how many days any discussion thread continued)

c. the least delay between messages. [-8-]

It could well be that students who reached Module 3 felt less need for X-citation because of these features which bring CMC closer to synchronous live discussion. If this is the case, the implications for course design are clear: it needs to be structured in small chunks, with very clear, bounded topics, and within short time-spans. How achievable such a structure is will depend on many factors relating to the subject matter of the course, and success in motivating students to honour the time commitments required. This point will be returned to in the conclusion below.

Finally, the data suggests yet another possible reason for the use or lack of X-citations. In the conference (Cafe) where students sent more personal non-course-related messages, there was so little X-citation that it was not included for the statistics of this paper. It appears as if X-citation occurs where more earnest consideration is being given to ideas, an intuitively reasonable conclusion.

We now turn to the development of X-citation conventions.

Conventions in Normal Language Use and CMC

It is well-accepted that conventional systems lie at the heart of language, and a considerable amount of linguistic research has been devoted to analysing them. Conventions develop in a variety of ways under different influences. The fundamental work by Lewis (1969) suggests they are initiated by a few people and then gradually accepted by the rest. In natural language development, linguistic conventions are rarely consciously known, let alone discussed. A problem in linguistics, as in all social science, is how to account for those influences on behaviour that lie beyond the actors' consciousness. But in the new medium of CMC we have a readily observable record of such a process, and the paper will briefly describe how the users in our sample struggled and negotiated with each other to develop suitable conventions of interaction.

In the technological context, a crucial factor is that pre-existing norms built into the software create conventions which users are compelled to live with. In e-mail and conferencing software, communication is pre-patterned in memorandum message style. Each message is preceded by headers indicating From whom it comes, To whom it is addressed, the Date, the Subject and perhaps other information. There are few options open to the user at this level. For instance, anonymous messages are not possible. It is not easy for educators to influence the conventions within which the online communication is constrained. But we ought to try, and later in this discussion I will make some proposals for software to assist with X-citation conventions. [-9-]

Within the given software limitations, then, users have to find ways of referring back to things previously said. But obviously, the reference methods used in oral discourse cannot just be imported into writing.

The development of X-citation conventions is also heavily influenced by a range of factors having to do with the production of written text. Until very recently, writers were used to leaving the actual production of texts to typesetters, printers, and others with typographic skills. Most writers learned a very narrow range of writing conventions beyond spelling, principally to do with personal and business letter-writing. Some, mostly academics, also learned some conventions to do with publishing, but someone else composed the final product.

In CMC discourse, writers must create the whole product themselves, taxing their skill in using the tools. Moreover, in learning how to create a CMC analogy of reference conventions, users have no sources of guidance; there are no ready manuals or other overt Help materials.

Of course, as in all language use, people have a vast range of intuitive competence gathered over many years of exposure to textual conventions. In developing X-citation conventions for themselves, our sample users were displaying just the kind of linguistic creativity that is fundamental to all language use.

Nonetheless, such creativity is not universal or perfect. For example, it can be readily shown (Norrish, 1990) that users can quickly and easily recognise the original contexts of random selections of written texts (newspaper, telephone directory, advertisement, etc.) but that, without special training, they cannot easily produce all types of text genres. Similarly, only some of our student developed satisfactory X-citation methods.

The use of X-citation was driven, as has been shown, by a perceived need to ensure that the reader had the same mental representation of the preceding dialogue as the writer. This perception can be linked to the now well-known maxims for successful communication (Grice, 1975). Interlocutors strive to be relevant and informative, but with maximum brevity and clarity. It is difficult to balance these goals. The maxim of brevity would urge CMC writers to avoid or abridge X-citations, but at the risk of loss of relevance or clarity. [-10-]

There is built-in relevance feature in most software, namely the option to send Reply messages that retain the same subject header as the message being replied to. The subject header thus becomes a visible link, even where the sender of the original message receives many intervening messages before this respondent's reply arrives. If users send a string of messages in this way, a "thread" (the computer conferencing term) is created.

Most software (like FirstClass) allows users to rearrange the messages on their screen so as to cluster all those that have the same subject header. Such threads can then be read cohesively in sequence. Experienced CMC users may assume their fellow discussants will do this. But it cannot be taken for granted, and it was evident on our course that writers used X-citation within their even within such apparently cohesive threads.

Figure 2: X-citations and threads shows that the majority of X-citations were in fact from within the same thread, and only a very small fraction of X-citations were from outside it. For example, in Module 1, only 1.8% of X-citations were from a second thread, and even fewer (0.4%) from a third thread. In other words, the pressure to be clear pushed users to reiterate (X-cite) parts of previous messages, even though the subject headers made the source of the reference easily recoverable and they were violating the maxim of brevity.

It is evident, therefore, that the perceived need to use X-citations is based, not on logic, but on users' feelings about clarity and relevance. None of the data in this sample shows any correlation between frequency of X-citation use and numbers of messages, word counts, or numbers of participants - as will be further demonstrated below.

This is an important finding, because it may mean that attempts to discover or to devise strict rules of X-citation behaviour would be doomed to failure. Of course, this is not surprising in the field of linguistics, where the discourse "rules" that have been discovered are, on the whole, as vague as Grice's maxims. What regularity there is in language seems not to apply very strictly or predictably to natural discourse.

The variability of X-citation use can be seen in the statistical data. Figure 3: Overall Messages, word count and X-citations show how X-citation varies at random at least in relation to the factors measured in the present study.

In Module 1 it appears to relate to the number of participants: the Plenary (50 participants) has X-citations in 30.9% out of the total of 97 messages and 11,048 own words used. The smaller group discussions (Apaloosa - Shire, with 8-9 members each and often with similar word counts to the Plenary), have much lower X-citation figures. For instance, Hanoverian has 10,200 own words used and 76 messages, but a 10.5% X-citations measure. [-11-]

But this is not borne out by other comparisons. For instance, the two groups, Hanoverian and Shetland, each with 17-18 members, 76 messages and reasonably close total word counts (10,200/13,921) have respectively 10.5% and 25% of messages with X-citation.

Nor is it a matter of word counts. Module 3 Plenary with 18,807 own words used has 29.5% X-citation use, while workshop 1 with 1608 own words used has 48.2%.

Further, a breakdown of X-citation use by individual students again demonstrates that numbers of messages or word counts do not correlate with X-citation use. There are examples such as a student who sent only 9 messages throughout the course, and wrote a total of 1989 words, but that total includes 44.4% of messages (450 words) using X-citations. By contrast, there are examples like the student who sent a total of 94 messages with a total of 18,288 words, of which 50% (3336 words) had X-citations.

In other words, it is not sheer volume which causes people to feel the need to cite in ways that they never do in conversation. It is clear that discourse needs are the reason, i.e., users' efforts to be clear and informative within the CMC context as they perceive it at the time of writing. But these needs are interpreted flexibly and variably by users themselves as the discussion develops.

When translating oral referencing conventions to written X-citations, users are faced with two different kinds of choice:

1. How much of the original message to cite or paraphrase

2. How to present the X-citation graphically.

Both may depend on what the software allows. In some systems, the Reply function delivers the full original message followed by a space for the response. In others, this is done by the Forward function, as in the version of FirstClass used on the course. Many of our students discovered that Forwarding was less work than copying and pasting the original message into a Reply. This led to an unfortunate complication. FirstClass, like many conferencing systems, has an optional thread facility that allows readers to see messages in sequence from the original through all Replies to it, but does not include Forwarded messages. So there were threads from which Forwarded messages were missing.

The various attempts students made to create effective X-citations fell into two categories. First, they experimented with the presentation. For instance, when they saw that FirstClass prefaces the Forwarded original message with " writes:..." they corrected it to " wrote/said:", followed by "me here:" or " writes:" and then their own response. [-12-]

Another presentation option in FirstClass is on-screen colour (though once messages are saved, the colours would only be seen by readers who have colour-sensitive word processors or printers). Two students, D. and J., conducted an interchange using different colours for cited words. Some students enthusiastically copied this method. One commented "colour makes the discussion more readable and more like a conversation because using colours to differentiate voices is a good way to follow the line of a conversation involving multiple individuals". Some, however, wrote comments like "When I read the entries with the colored text in scroll, I find it rather distracting."

Second, the students experimented with abbreviating an original message. Sometimes they just cut it, sometimes they paraphrased it. There is very clearly a skill in this, and students' choice of method became more complex as the discussion grew. After one person had cited and responded, the next person might cite the original and the first response before adding another comment.

Commentators also found they had to decide whether to respond to each message individually straight after reading it, or to read a series at a time and then respond to all the messages in that series. In the latter case the result can be a very complex web of criss-crossing references. Their methods varied from citing the whole preceding set through citing only part, to giving some paraphrased version or summary.

One student wrote about both the use of colour and of multiple X-citations: "I kept finding it incredibly irritating having these endless repeats of messages, all beautifully arrayed in red, blue and black, but actually with each message it took me a few minutes to work out which was the new contribution and what was the repeat of before, especially when the new comment was put at the bottom of the repeat. I have the same problem with my daily e-mail - having screenfuls of previous message before finding the new comment. It's not so bad when the answer comes before the repeat of the original (although that might seem silly) because at least then you have an easier choice about reading it or not."

The group reached no consensus either about quantity of X-citation or methods of presentation in colour. This investigation showed no consistency, either between the number of words that were cited on average per messages or the number of citings per message or the number of sources, ie original messages, from which X-citations came. See Figure 4: Word counts and sources of X-citations. [-13-]

About half-way through the course, I, as tutor, suggested that the problem might be related to the awkwardness of the FirstClass software which presents and saves messages separated in discrete windows and does not allow a series of messages to be read as one continuous prose stream. I proposed that the group try out a way of mimicking what happens in F2F conversation, i.e., using Reply style only:

1. Everyone should use Reply without any X-citations; subject headers would indicate which message a Reply related to.

2. If people sorted their messages by subject header, there might be no need for X-citations.

The students did not, as a body, take up the proposal, but one student, L., used it to conduct a small experiment. He was reacting against the use of colour to signal X-citations.

He provided 17 experienced e-mail users with three different conversations, A, B, C, each of which was presented in two different ways: one with messages using X-citations in colour, and one following the proposed Reply style above. The subjects were asked to join in a conversation if the method of threading appealed to them. The results show that the small sample had a clear preference for the Reply style i.e. doing without X-citations.


Joined the "reply" style

Joined the colour
X-citation style

Joined Both

Conversation A 7 2 2
Conversation B 9 0 5
Conversation C 6 1 2

This may, however, be a situation where experimental results do not mirror natural behaviour. The experiment used very short conversations and the results may be representative of situations like that referred to above regarding the low use of X-citations in many threads of Module 3. However, the data of the present paper, and other experience, strongly suggests that spontaneously occurring, ongoing, longer discussions are likely to be too complex to follow without X-citation conventions of some kind. [-14-]

During the course, other solutions were proposed to differentiate voices where X-citations were needed:

  1. inserting one's initials before each comment e.g
    OB: I don't agree with this at all, sorry.

    PW: But, Owen, Anita does have a point.

  2. Inserting a picture or cartoon of the speaker at the start of each of his/her comments.
  3. Using quotation marks and the full name of each contributor.

But students replied, inter alia, that they could not remember initials and predicted that in any case it was too much to expect everyone in a large group to keep consistently to the same principles of X-citation. Indeed one student pointed out that she could not even rely on clear use of subject headers, which, in her view were the key to CMC coherence. She gave, as part of her "wish list", that people would make their subject headers regularly reflect what they wrote about, and not "go off at tangents under one subject header".

This is consistent with all investigation of normal discourse: in spite of some patterns of genre and many rhetorical conventions that have developed for different contexts of use, procedural discipline goes against the grain of ordinary language behaviour. CMC discussants are as varied, creative and adaptable in their written practice as in normal conversation.

This is why it would be useful if software producers could improve the functions they offer. It would not be technically difficult to incorporate features to facilitate X-citation. For example a system in which, among the headers of every message, there was a new kind of header (perhaps called "X-references") which would tell the reader what other messages were linked to to the current one, e.g., by similar words in the text, by common X-citations, etc. Such a header would have to be automatically up-dated as the discussion progressed. The reader could then opt to view references at will. The danger with this method is that the new header could soon become unusably long.

Better might be a system whereby, if a person used a phrase that linked with a previous message, an automatic hyperlink was created giving the reader the opportunity of "flicking back" as we do with books. Or again, an automatically performed search might find and display similar phrases from previous messages. As soon as such functions exist in the software, users will develop ways of benefiting from them. [-15-]


In F2F situations, one type of solution to the difficulties of discussion among large numbers of people is seen in the highly structured committee context, where convention gives a chairperson the right to enforce an agenda and hold discussants to prescribed turns and roles. In CMC, it is only possible for a chairperson to manipulate discussion after messages have been posted, e.g. by removing some or rearranging their order. But this is merely to repair the interactions, not guide them let alone control them (except to the extent that the discussants might learn a mode of procedure).

Charles Jennings (one of the tutors on the course) pointed out that CMC systems "allow us to step outside the single linear conversational model we are bound to" in normal talk. "If we accept this, ... we will see CMC as a distinct mode of communication in its own right, rather than merely a substitute for the familiar F2F talk ... We will then stop trying to replicate F2F events" and develop conventions appropriate to our needs in CMC. He was referring, in part, to the obvious CMC benefits, such as the ability to run several discussions at the same time, the possibility of everyone making contributions without interruption, and so on. One way of assisting this is to amend the software, as suggested above.

But we can also learn from F2F conventions. The data presented has already shown that CMC users can conduct discussion without many X-citations, if the context of the discussion is close to a F2F framework, i.e., if it is well bounded and messages are sent without too much lapse of time.

On courses other than the one discussed here, I use a design that attempts to approach F2F success by pre-structuring the discussions (Pincas 1995, 1998). The whole period of the course is cut into short - one week time frames, students are divided into small groups of no more than 12, within which every individual has a separate, brief, task to respond to. The participants are provided with a detailed model of the correct subject headers to use for their messages: all initial task answers have simply the task number as header. All subsequent comments on a task have the same number plus "Comm 1" for the first comment, and "Comm 2" etc. for subsequent comments. [-16-]

As a result, there is virtually no X-citation, and no confusion about reference. I have no indications that students are confused. On the contrary, judging by their performance, they seem to have a quite satisfactory mental representation of the course. The following are some extracted examples taken from a currently running Masters degree by e-mail for language teachers (names have been removed):

Student 1
Subject: AL2 - 4 Comm2
I would like to say that I enjoyed reading the members' messages about the use of recorded voice and recorded texts (D. K.,, H. and S.) and that I also appreciated S.'s task answer and N.'s comment on it. I just add two comments.......

Student 2
Subject: AL2 - 4: Comment
I read with interest F.'s response to Task 4, regarding language labs. I remember as a high school student using a language lab ...

Student 3
Subject:: AL2 - 7 Comm4
Well, I just had to respond to this one! I am the "APE" that D. mentioned. Honestly, my on campus job when I was in college...

These students are highly motivated to complete their Masters, and they know that regular e-mail attendance is a condition of passing. They therefore comply with the arrangements, read and respond to messages frequently, and use correct headers as instructed.

By contrast, the Certificate in Online Education and Training consists of very busy academics interested in experiencing CMC, but not able to devote very much time to the course. They have never responded well to tight structuring because it necessitates regular work to a timetable. They prefer to engage in wide-ranging open-ended discussion, with the resulting difficulties described in this paper.

I would urge online course designers to consider carefully the benefits of structured interactions over free messaging.


I am indebted to a former student (Labadie, 1997) for the statistics in this paper, which are taken from his research into the use of CMC during the course on which this paper is based. [-17-]


Boden, D. (1994).The Business of Talk. Cambridge: Polity Press. Catt, R. & Eke, J. (1995). Classroom Talk in Higher Education: Enabling Learning through a Reflective Analysis of Practice.Innovations in Education & Training International, 32,4. (pp. 362-369).

Grice, H.P. (1975). Logic and Conversation, In P.Cole & J.Morgan (Eds.), Syntax and Semantics 3: Speech Acts. New York: Academic Press.

Halliday, M.A.K. & Hassan, R. (1976).Cohesion in English. London: Longman.

Halliday, M.A.K. & Hassan, R. 1990).Language Context and Text. London: Oxford University Press.

Harasim, L. (1989). Online Education: A New Dimension, In R. Mason, & A.Kaye (Eds.),Mindweave. (pp. 50-62). London: Pergamon.

Howell-Richardson, C. (1995). Interaction across computer-conferencing, In R. Howard & I. McGrath (Eds.),Distance education for language teachers. Avon: Multilingual Matters Ltd.

Howell-Richardson, C. (1996). Read-time conversation on the Internet?, In G. Motterham, G. Walsh & R. West (Eds.),Distance education for language teachers: 2nd symposium. Manchester: School of Education, University of Manchester.

Howell-Richardson, C. & Mellar, H. (1996). A methodology for the analysis of patterns of participation within computer mediated communication courses.Instructional Science 1. (pp.47-69).

Kaye, A. R. (Ed.) (1991).Collaborative Learning through Computer Conferencing. Berlin: Springer.

Labadie, C. (1997). Can handshaking and citations help CMC fit into the busy life of medical doctors?, Unpublished essay, Institute of Education, University of London, London.

Laurillard, D. (1993).Rethinking University Teaching. London: Routledge.

Lewis, D.K. (1969).Convention: A philosophical study. Cambridge: Harvard University Press. [-18-]

McConnell, D. (1994).Implementing Computer Supported Cooperative Learning. London: Kogan Page.

Mason, R.D. (1993).Computer Conferencing: the Last Word. Victoria: Beach Holme.

Mason, R. & Kaye, A. (Eds.) (1990).Mindweave. London: Pergamon.

Norrish, P. (1990).The Graphic Translation of Text. Unpublished doctoral dissertation, University of Reading, (p.16).

Paulsen, M.F. The Online Report on Pedagogical Techniques for Computer-Mediated Communication. Available: http://www.nki.no/~morten/

Pincas, A.(1995a). Analysis of face to face and computer conferencing interactions in university teaching" In C. O'Hagan (Ed.)Empowering Learners and Teachers through Technology 90 (pp. 109-115) Birmingham: SEDA (Staff & Educational Development Association).

Pincas, A.(1995b). Brief case studies of two university computer conferencing courses, In F. Percival, R. Land & D. Edgar-Neville,Aspects of Education & Training Technology XXVIII (pp.216-222) London: Kogan Page.

Pincas, A.(1995c). Understanding Networked Learning communities, In P. Held & W. Kugemann (Eds),Telematics for Education and Training (pp.277-281) Oxford: IOS Press, and Tokyo: Ohmsha Ltd..

Pincas, A.(1995d). Solving the problems of learner collaboration in asynchronous computer conferencing courses, In R. Cornell, & K. Murphy (Eds.),From Smoke Signals to Satellite II (pp.115-121) Orlando: University of Central Florida.

Pincas, A.(1998). New Media, Homogenous Knowledge and Education, Continuing Professional Development, 2, 2 Available: http://www.openhouse.org.uk/virtual-university-press/

Pincas, A.(1998). Successful online course design: Virtual frameworks for discourse construction EducationalTechnology & Society, 1, Available: http://zeus.gmd.de/ifets/periodical/ [-20-]

Porter, M.(1990).The Mode of Information. New York: Polity Press. Robinson, B. (1993). Communicating through Computers in Classrooms, In P.C.

Scrimshaw (Ed.).Language, Classrooms and Computers. London: Routledge.

Rowntree, D. (1995). Teaching and learning online: a correspondence education for the 21st century.British Journal of Educational Technology, 26, 31-39.

Saville-Troike, M. (1982).The Ethnography of Communication. Oxford: Blackwell

About the Author

Anita Pincas, Lecturer in Education. Course Tutor for the MA TESOL by Email. She has previously taught and trained teachers of English in the education departments of Sydney University Australia, Columbia University Teachers' College, New York, and the Hebrew University of Jerusalem. Her special interests are the teaching of writing, pedagogic grammar, and the use of computer-mediated communication for distance education. She has pioneered the world's first MA TESOL by Internet, making the Institute of Education a leader in this field. She also initiated and runs the Certificate in Online Education and Training, a course for educators who wish to experience and discuss methods of course delivery by electronic communications.

© Copyright rests with authors. Please cite TESL-EJ appropriately.

Editor's Note: Dashed numbers in square brackets indicate the end of each page for purposes of citation.

Return to Table of Contents Return to Top Return to Main Page



Figure 1: Overall use of X-citation

1 X-citation per message 12.1% 12.1% 5.6% 24% 36%
2 X-citations per message 1% 4% 1% 8% 10%
3 X-citations per message 6% 1% 1% 2% 4%
4+ X-citations per message 1% 1% 3.5% 3% 4%
PARTICIPATION IN 1 Activity 2 Activities 3 Activities 4 Activities All 5 Activities

Back to the text


Figure 2: X-citations and threads.

  Module 1 Module 2 Module 3 Module 4 Workshops
Total number of messages

490 361 396 86 140
Messages with 1+ X-citation(s) 16.5% 34.1% 20.5% 17.4% 37.1%
Messages with X-citation(s) from 2 threads 1.8% 3.6% 2.3% - 3.6%
Messages with X-citation(s) from 3 threads 0.4% 1.1% 0.3% - -
Messages with X-citation(s) from 3+ threads - - - - -

Back to the text


Figure 3: Overall Messages, word count and X-citations.

Modules Messages Word counts Messages with X-citations
  user messages messages that are readings Total own words readings cited wds. Total wds. nos %
Module 1


97 0 97 11048 0 2205 13253 30 30.9
Readings 1 6 7 45 21390 0 21435 0 0
Apaloosa 48 0 48 7480 0 836 8316 9 18.7
Arab 65 0 65 9182 0 83 9265 3 4.6
Hanoverian 76 0 76 10200 0 459 10659 8 10.5
Percheron 67 0 67 13549 0 436 13985 5 7.4
Shetland 76 0 76 13921 0 1816 15737 19 25
Shire 60 0 60 10932 0 2031 12963 12 20
Module 2


99 0 99 14507 0 3585 18092 35 35.3
Readings 8 11 19 1625 67381 0 69006 0 0
Group A 104 0 104 20099 0 5953 26052 50 48
Group B 79 0 79 22569 0 14679 37248 35 44.3
Group C 45 0 45 12490 0 245 12735 5 11.1
Visiting tutor 26 3 29 4631 2477 1003 8111 6 23
Module 3


149 0 149 18807 0 3195 22002 44 29.5
Readings 0 12 12 0 52183 0 52183 0 0
Case Studies 13 0 13 9274 0 11 9285 1 7.6
Discussion 206 0 206 34009 0 1354 35363 30 14.5
Visiting tutor 29 2 31 3264 14914 281 18459 7 24.1
Module 4


13 0 13 2036 0 110 2146 3 23%
Readings 15 26 41 1367 82051 69 83487 2 13.3
Notes 27 1 28 3545 1661 366 5572 6 22.2
Tasks 31 1 32 4467 1045 229 5741 6 19.3
Workshop 1 85 0 85 16086 0 6398 22484 41 48.2
Workshop 2 23 0 23 3343 0 378 3721 7 30.4
Workshop 3 32 0 32 9183 0 261 9444 5 15.6
Total 1474 62 1536 257659 243102 45983 546744 369 24

Back to the text




Figure 4: Word counts and sources of X-citations

Messages containing 1+ citation Module 1 Module 2 Module 3 Module 4 Workshops
Average number of cited words per message 90.7 199.4 59.6 46.0 129.4
Average number of cited items per message 1.72 1.91 1.23 1.13 1.85
Average number of source messages per message 1.15 1.41 1.05 1.13 1.10

Back to the text